Spark-X2.6-1.7B Benchmarks, Specifications & Availability

Spark-X2.6-1.7B

Explore Spark-X2.6-1.7B from SparkLLM: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.

Published specifications

Provider: SparkLLM
Access: Open weights
License: Apache-2.0
Context window: 1.05M
Total parameters: 1.71B
Active parameters: Not published
Released: 2026-10-19
Modalities: text
Family: Spark-X2.6

Model card & announcement · Website

Model notes

SparkLLM (XHToken) open on-device model, architecture spark2_5, thinking levels none, low and high. Its card server is configured for 1,048,576 tokens of context; that is a setting, not a validated accuracy limit. SWebBench and PAgentBench are in-house. Peer rows follow the card figure: starred values come from other cards and are not stored; Spark-X2.5, Qwen3.5, MiniCPM5 and LFM2.5 peers are unregistered and get no stubs. No OpenRouter route on 2 November 2026. 1,707,657,216 safetensors parameters. Announced date is the repository creation date; the first SparkLLM post found is dated 27 October. Older omission, surfaced with the 4B on 1 November 2026.

Compare Spark-X2.6-1.7B with other models →Compare subscription plans Explore data coverage

Pricing · OpenRouter

No OpenRouter pricing is recorded for this model. Missing rates are not free.

Official / vendor benchmarks

Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.

Official / vendor headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
Average of 30 benchmarks (SparkLLM)	49.6 scorehigh effort	All 1 recorded result & sources 49.6 score · raw 49.6 score Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “Average” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
BFCL v4 (overall)	55.2%high effort	All 1 recorded result & sources 55.2% · raw 55.2 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “BFCL-V4” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
τ²-bench (Tau-2)	85.3%high effort	All 1 recorded result & sources 85.3% · raw 85.3 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “τ²-bench” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
τ³-bench Banking	21.3%high effort	All 1 recorded result & sources 21.3% · raw 21.3 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “τ³-bench Banking” in the card figure. Self-reported; run date unpublished. The 4B figure labels this τ³-bench; the 1.7B figure labels it τ³-bench Banking with the same 4B reference value. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
VitaBench 2.0 (SparkLLM)	22.6%high effort	All 1 recorded result & sources 22.6% · raw 22.6 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “VitaBench2.0” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
MCP Atlas	41.8%high effort	All 1 recorded result & sources 41.8% · raw 41.8 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “MCP-Atlas” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
EnterpriseOps-Gym (SparkLLM; run unspecified)	26.1%high effort	All 1 recorded result & sources 26.1% · raw 26.1 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “EnterpriseOps-Gym” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
ClawEval-EN (SparkLLM)	66.9%high effort	All 1 recorded result & sources 66.9% · raw 66.9 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “ClawEval-EN” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
BrowseComp	41.7%high effort	All 1 recorded result & sources 41.7% · raw 41.7 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “BrowseComp” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
SWebBench (SparkLLM in-house)	53.9%high effort	All 1 recorded result & sources 53.9% · raw 53.9 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “SWebBench” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
PAgentBench (SparkLLM in-house)	58.8%high effort	All 1 recorded result & sources 58.8% · raw 58.8 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “PAgentBench” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
SWE-bench Pro (revision unspecified)	26.7%high effort	All 1 recorded result & sources 26.7% · raw 26.7 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “SWE Pro” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
SWE-bench Verified	59.8%high effort	All 1 recorded result & sources 59.8% · raw 59.8 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “SWE Verified” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
SWE-bench Multilingual	57.3%high effort	All 1 recorded result & sources 57.3% · raw 57.3 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “SWE Multilingual” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
SciCode (SparkLLM; setup unspecified)	20.1%high effort	All 1 recorded result & sources 20.1% · raw 20.1 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “SciCode” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
Gaokao Math 2026 (SparkLLM, five papers out of 150)	118.9 pointshigh effort	All 1 recorded result & sources 118.9 points · raw 118.9 points Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “Gaokao 2026” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
AIME 2026	70.1%high effort	All 1 recorded result & sources 70.1% · raw 70.1 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “AIME 2026” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
HMMT February 2026	48.8%high effort	All 1 recorded result & sources 48.8% · raw 48.8 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “HMMT Feb 26” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
IMO-AnswerBench (SparkLLM)	47.1%high effort	All 1 recorded result & sources 47.1% · raw 47.1 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “IMO-AnswerBench” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
AA-LCR v1.1	48%high effort	All 1 recorded result & sources 48% · raw 48 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “AA-LCRv1.1” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
LongBench v2	40.4%high effort	All 1 recorded result & sources 40.4% · raw 40.4 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “LongBench v2” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
GraphWalks BFS F1 (256K context)	85.1 f1high effort	All 1 recorded result & sources 85.1 f1 · raw 85.1 f1 Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “Graphwalks bfs 256k f1” in the card figure. Self-reported; run date unpublished. F1 on a 0–100 scale, as the figure prints it. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
GraphWalks BFS F1 (1M context)	25.7 f1high effort	All 1 recorded result & sources 25.7 f1 · raw 25.7 f1 Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “Graphwalks bfs 1m f1” in the card figure. Self-reported; run date unpublished. F1 on a 0–100 scale, as the figure prints it. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
MRCR 256K–512K	56.1%high effort	All 1 recorded result & sources 56.1% · raw 56.1 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “Openai_MRCR v2 256k-512k” in the card figure. Self-reported; run date unpublished. Source label OpenAI MRCR v2; needle count unstated on the card. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
MRCR 512K–1M	19.9%high effort	All 1 recorded result & sources 19.9% · raw 19.9 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “Openai_MRCR v2 512k-1M” in the card figure. Self-reported; run date unpublished. Source label OpenAI MRCR v2; needle count unstated on the card. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
IFEval	89.4%high effort	All 1 recorded result & sources 89.4% · raw 89.4 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “IFEval” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
IFBench	65.5%high effort	All 1 recorded result & sources 65.5% · raw 65.5 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “IFBench” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
Humanity's Last Exam (no tools)	8.7%high effort	All 1 recorded result & sources 8.7% · raw 8.7 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “HLE(w/o tools)” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
Humanity's Last Exam (with tools)	19.9%high effort	All 1 recorded result & sources 19.9% · raw 19.9 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “HLE(w tools)” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
GPQA (SparkLLM; subset unstated)	51.3%high effort	All 1 recorded result & sources 51.3% · raw 51.3 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “GPQA” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg
GPQA Diamond	56.1%high effort	All 1 recorded result & sources 56.1% · raw 56.1 % Headline · high effort · Own vendor Source/record date: 2026-10-19 Harness: SparkLLM own run at thinking high per its card, temperature 1.0, top_p 0.95; exact label “GPQA Diamond” in the card figure. Self-reported; run date unpublished. https://huggingface.co/XHToken/Spark-X2.6-1.7B/resolve/e6406af83988f26e87ea92455f31db3f6476131b/images/all_benchmark.svg

Independent evaluators

Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.

No independent evaluator results are recorded for this model.

Read how we select and source scores or the comparison guide.